关键词: 网络流量, 相空间重构, 相关向量机, 组合核函数, 布谷鸟算法

Abstract: In order to obtain good forecasting results and describe the change rules network flow, a novel network flow forecasting model is proposed based on Hybrid kernels Relevance Vector Machine and Cuckoo Search algorithm（CS-HRVM）. Firstly, the learning samples are obtained by using phase reconstruction theory, and the hybrid kernels function is used to establish the relevance vector machine, and then the learning samples are input into relevance vector machine to train, and the cuckoo searching algorithm is used to optimize the parameters of model and build the model of network flow, finally, the simulation experiments are carried out to test the performance of the model. The results show that CS-HRVM has obtained higher forecasting accuracy compared with other network flow forecasting model, and can forecast accurately network flow which includes noise.

Key words: network flow, phase space reconstruction, relevance vector machine, hybrid kernel function, cuckoo search algorithm